IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 37, NO.

3, MAY 2022 1979

Distributed Finite-Time Event-Triggered Frequency

and Voltage Control of AC Microgrids
Jeewon Choi , Student Member, IEEE, Seyed Iman Habibi , Graduate Student Member, IEEE,
and Ali Bidram , Senior Member, IEEE

Abstract—This paper proposes a finite-time event-triggered sec- in grid-tied mode [1]. Traditionally, a centralized controller
ondary frequency and voltage control for islanded AC microgrids collects information from every DG in the secondary control
(MGs) in a distributed fashion. The proposed control strategy can layer, containing risk of being a single point of failure. As
effectively perform frequency restoration and voltage regulations,
while sharing the active and reactive power among the distributed opposed to the centralized control, the distributed controllers
generators (DGs) based on their power ratings. The finite-time need to interchange information only with their neighboring
control enables a system to reach consensus in a finite period of time controllers. Such distributed secondary control architecture can
enhanced from the asymptotic convergence. The event-triggered increase the MG system reliability and scalability, as well as im-
communication is utilized to reduce the communication burden proving the computational efficiency [3]–[6]. The decentralized
among the DG controllers by transmitting data among DGs if
an event-triggering condition is satisfied. The performance of the secondary control of AC microgrids has been proposed in [7].
proposed finite-time event-triggered frequency control is verified Even though the decentralized secondary control obviates the
utilizing a hardware-in-the-loop experimental testbed which sim- requirement for a communication network, it still relies on the
ulates an AC MG in Opal-RT. primary control level and it is not effective for some secondary
Index Terms—Distributed control, event-triggered control, control applications like regulating the voltage of a critical
finite-time control, islanded AC microgrid, multi-agent systems, remote bus.
primary control, secondary control. The distributed secondary control scheme has been further
advanced by adopting the various control techniques. The finite-
I. INTRODUCTION time control strategy enables a system to converge to the steady
state in a finite amount of time, enhanced from the asymptotic
ICROGRIDS (MGs) as critical components of modern
M power systems play a significant role in increasing the
reliability and resilience while facilitating the integration of
convergence time [8]–[10]. With finite-time control, the micro-
grid’s voltage and frequency will be converged faster to their
nominal values and experience much less transients. Both of
sustainable energy resources. The operation of MGs in both grid-
these features are very important for the microgrid critical loads.
tied and islanded modes is the key factor to enhance the power
According to [11], voltage and frequency transients can result
grid resilience during extreme events like natural disasters. As
in machinery malfunction as well as microgrid’s equipment and
a critical component to obtain reliable and stable operations in
components damage and failure over time. The frameworks
both modes, the hierarchically structured MG control system
for the frequency/voltage restoration and active/reactive power
has been surveyed in the literature [1], [2]. The hierarchical
sharing of an islanded AC MG using the finite-time control
control system is composed of the primary, secondary, and
method have been introduced in the literature [12]–[14]. The
tertiary control layers [1], [2]. The primary control is locally
frameworks introduced in [12]–[15] validates the effectiveness
deployed by the distributed generators (DGs) to stabilize the sys-
of finite-time secondary control in MG system, as well as
tem frequency/voltage and to satisfy the power sharing among
presenting its robustness against the disturbances such as load
DGs after the MG is disconnected from the upstream power
changes.
network. The secondary control restores the frequency/voltage
Even though the distributed finite-time control techniques
deviations caused by the primary control. The tertiary control
improve the frequency restoration speed of MGs, the DG local
regulates the power flow between the main grid and the MG
controllers continuously transfer their local information to other
Manuscript received January 22, 2021; revised May 11, 2021 and July 8, DGs which incurs unnecessary burden on the communication
2021; accepted August 22, 2021. Date of publication September 8, 2021; date system. Recently, the event-triggered communication has been
of current version April 19, 2022. This work was supported by the National applied to the multi-agent systems (MAS) [16]–[18] to tackle
Science Foundation EPSCoR Program under Grant #OIA-1757207. Paper no.
TPWRS-00120-2021. (Corresponding author: Jeewon Choi.) this challenge. As apposed to the periodic communication be-
Jeewon Choi is with the Department of Mechanical Engineering, University of tween the agents in a conventional MAS, the event-triggered
New Mexico, Albuquerque, NM 87131 USA (e-mail: [email protected]). method activates the communication only if an event-triggering
Seyed Iman Habibi and Ali Bidram are with the Department of Electrical and
Computer Engineering, University of New Mexico, Albuquerque, NM 87131 condition is satisfied, which in turn leads to the communica-
USA (e-mail: [email protected]; [email protected]). tion burden reduction. The event-triggered control has been
Color versions of one or more figures in this article are available at applied to the conventional MG distributed control systems in
https://doi.org/10.1109/TPWRS.2021.3110263.
Digital Object Identifier 10.1109/TPWRS.2021.3110263 order to decrease the communication network traffic among

0885-8950 © 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 1980 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 37, NO. 3, MAY 2022

DGs in [19]–[26]. The complete framework for the secondary

frequency/voltage restoration and active/reactive power sharing
of AC MGs using the event-triggered communication is pre-
sented by [20]. The effectiveness of the event-triggered control
in the distributed secondary control under the various scenarios,
such as a network delay, noise in the event-triggering, load
changes, and plug and play functionality of DG, is evaluated
in [19], [21]–[23].
The aforementioned event-triggered strategies can only pro-
vide the asymptotic consensus of MAS. In practical situations
however, it is desired to obtain the finite-time consensus. On
that account, numerous tactics to attain a finite-time consensus
Fig. 1. Block diagram of ith inverter-based DG.
under the event-triggered communication in a general MAS
are introduced in the literature [27]–[34]; however, the event-
triggering mechanisms designed in [27]–[33] either require a
continuous transmit of information from the neighbors or cannot and voltage through the local frequency and voltage droop
guarantee the exclusion of Zeno behavior. In order to pursue a controllers, that is the primary control layer. In Fig. 1, the droop
finite-time consensus with an event-triggering communication controllers are implemented in the power controller block. The
that excludess a Zeno behavior, Du et al. proposed a finite-time roles of the primary control layer are to stabilize the system
event-triggered consensus (FETC) scheme in [34]. frequency/voltage subsequent to an islanding event as well as to
Although FETC has been proposed in the literature for a share the active/reactive power among the DGs based on their
general MAS, the FETC of the MG distributed control system power ratings [2]. The primary control, however, results in the
has not been addressed in the literature. The existing distributed frequency/voltage drop from the nominal operating values. The
secondary control techniques either address the event-triggered secondary control is implemented in order to restore the MG’s
operation or the finite-time consensus. A comprehensive dis- frequency/voltage to their nominal values. In this paper, the
tributed secondary control approach that accommodates a finite- secondary frequency/voltage restorations and the active/reactive
time control response while obviating the requirements for a power sharings among the grid-forming inverters are of concern.
continuous communication is missing in the literature. There- The primary control is locally implemented in each inverter’s
fore, this paper creates a unified distributed secondary control controller [2]. The angular frequency ωi and the reference value
∗
for MGs that accommodates both the event-triggered communi- for the internal voltage controller vo,magi of DG i, are determined
cation and finite-time response features. To this end, this paper by the following droop characteristics
makes the following contributions:
ωi = ωni − mP i Pi (1)
1) A FETC scheme for the MG’s distributed secondary fre-
∗
quency and voltage control is created to achieve both vo,magi = vni − nQi Qi (2)
consensus in finite time and lower communication burden
in the MG communication system. where mP i and nQi are the active and reactive power droop
2) The exclusion of Zeno behavior in the proposed distributed coefficients, respectively, that relate to the power ratings of a
frequency and voltage control is investigated. DG. ωni and vni are the references for primary control. The DG’s
3) A strategy is proposed to effectively accommodate an internal voltage and current controllers converge the DG’s ter-
∗
islanded MG’s frequency and voltage control subsequent minal voltage magnitude vo,magi to vo,magi . The active/reactive
to load changes as well as energization or outage of DGs. power are distributed among the DGs by the primary controller
The remaining of this paper includes the following sections: proportional to their active/reactive power ratings, Pmax,i and
In Section II, the primary and the secondary control levels are Qmax,i , respectively. To this end, the active/reactive power ratios
introduced. The distributed secondary frequency and voltage of each DG are defined as Pi /Pmax,i and Qi /Qmax,i , respec-
restoration framework based on the event-triggered finite-time tively. The primary control ensures that the active/reactive power
control is proposed in Section III. The validation of the pro- ratios satisfy
posed secondary control through a hardware-in-the-loop (HIL) P1 PN
simulation is presented in Section IV. Section V concludes the = ··· = (3)
Pmax,1 Pmax,N
paper.
Q1 QN
= ··· = . (4)
Qmax,1 Qmax,N
II. INVERTER-BASED MGS
Since the droop coefficients are selected based on the power
This paper assumes that the AC MG is composed of inverter-
ratings, (3)–(4) can be equivalently expressed as
based DGs. The block diagram of an inverter-based DG is
depicted in Fig. 1. An inverter-based DG consists of the primary mP 1 P 1 = · · · = m P N P N (5)
DC power source, inverter bridge, and nested voltage, current,
and power controllers. DGs are able to control the MG frequency nQ1 Q1 = · · · = nQN QN . (6)

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 CHOI et al.: DISTRIBUTED FINITE-TIME EVENT-TRIGGERED FREQUENCY AND VOLTAGE CONTROL OF AC MICROGRIDS 1981

Fig. 2. Schematic of the proposed distributed FETC secondary frequency and voltage control.

It should be noted that once the DGs’ active/reactive powers sat- 0. The neighbors of node i are defined by a set of nodes that have
isfy (5)–(6), the angular frequencies ωi and voltage magnitudes a link to vi and can be written as Ni = {vj |(vj , vi ) ∈ E}. The in-
vo,magi , are settled into the values that are smaller than the nom- degree matrix of a graph is defined by D = diag{di } ∈ RN ×N

inal values. To restore MG’s frequency and voltage back to the where di = N j=1 aij . The graph Laplacian matrix is calculated
nominal values, the secondary frequency and voltage controls by L = D − A [36].
adjust ωni and vni of each DG in (1) and (2), respectively. This In [34], the following control input is proposed in order to
process will be discussed in Section III. achieve the finite-time consensus without continuous communi-
cation among agents
III. FETC-BASED SECONDARY FREQUENCY CONTROL
ui (t) = c1 Kzi (t) + c2 Ksig(R(zi (t) + ei (t)))p (8)
One of the objectives of an islanded MG control in the
secondary level is to synchronize the angular frequencies of where sig(x)p = sgn(x)|x|p with 0 < p ≤ 0.5, and sgn(·) is
DGs, ωi ’s, to the nominal angular frequency of the system, the signum function. c1 and c2 are positive real numbers, R ∈
ωref . At the same time, it needs to allocate DGs’ active powers Rn×n is a positive-definite matrix, and K = −B  R ∈ Rm×n .
based on their ratings according to (5). The schematic of the The estimate error ei (t) and the local tracking error zi (t) for ith
FETC-based secondary frequency control proposed is illustrated agent are defined as
in Fig. 2. The secondary controller calculates the primary control i
input using its local and neighbors’ information. The secondary ei (t) = xi (t) − eA(t−tk ) xi (tik ) (9)
controllers share information via communication graph, when  j
aij (eA(t−tk ) xi (tik ) − eA(t−tk ) xj (tjk ))
i
the event-triggering condition is satisfied. The details of the zi (t) = (10)
secondary frequency controller using the FETC strategy are j∈Ni
elaborated as follows.
where tik is the latest communication triggered time of ith agent
determined when the event-triggering condition fi (t) ≥ 0 is
A. FETC Problem for a General Linear MAS
met. To obtain the finite-time consensus, the following event-
Using input-output feedback linearization, the nonlinear dy- triggering function is proposed in [34]
namics of DGs are transformed to identical linear dynamics.
This process has been explained in details in [35]. Based on the fi (t, ei (t), zi (t), ϕi (t)) = α ei (t)2 + β ei (t)2p
dependence of DGs’ operating frequency and terminal voltage
+ γ zi (t)2p − χε1i |ϕi (t)|2ψ
magnitude on the active and reactive power droop reference val-
(11)
ues, feedback linearization leads to a first-order linear tracking
synchronization problem for a MAS [4]. Consider a MAS of ϕ̇i (t) = −ε2i sig(ϕi (t))ψ (12)
which ith agent’s dynamics is given by
where αei (t)2 + βei (t)2p + γzi (t)2p is the error term
ẋi (t) = Axi (t) + Bui (t). (7) and χε1i |ϕi (t)|2ψ is the event-triggering threshold. The trig-
gering condition fi (t) ≥ 0 indicates that the communication
where xi (t) ∈ Rn is the state variable vector, ui (t) ∈ Rm is
is triggered when the error term exceeds the threshold, i.e.,
the system input vector, A ∈ Rn×n is the state matrix and B ∈
agent i broadcasts its updated state xi (tik ) to its neighbors.
Rn×m is the input matrix. DGs can communicate with each
Unlike the asymptotic triggering thresholds introduced in the
other through a communication network which can be modeled
literature [16]–[23], [26], note that the threshold ϕi (t) developed
by a graph G = (V, E), where V = {v1 , . . . , vN } is a set of N
by (12) is a positive real number that monotonically decreases to
vertices or nodes, and E ⊂ V × V is a set of links. The weights
 zero in finite time [34]. The term ei (t)2p also helps obtaining
aij of the edges (vj , vi ) form an adjacency matrix A = aij ∈ the finite-time consensus [34]. α, β, and γ are positive real
RN ×N of graph G. If (vj , vi ) ∈ E, then aij > 0, otherwise aij = numbers greater than certain criteria, ε2i > ε1i > 0, χ > 0, and

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 1982 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 37, NO. 3, MAY 2022

0 < ψ ≤ p. The FETC strategy introduced can be summarized B. FETC Problem for a Secondary Frequency Control of MG
as Lemma 1.
Differentiating the droop characteristic in (1) yields
Lemma 1 ([34]): For the MAS given in (7), assume
(A, B) is controllable and the communication graph G is ω̇ni (t) = ω̇i (t) + mP i Ṗi (t). (13)
strongly connected. With the control law (8) and the event-
Define the auxiliary frequency and active power control inputs
triggering function (11), the system can reach consensus in
uωi and uP i , respectively, as
finite-time for all initial conditions if the following condition
is satisfied: ( θc11 + θc22 )λmax (RBB  R) + θ23 < −λmax (Ξ ⊗ ω̇i (t) = uωi (t) (14)
S) where θ1 , θ2 , , θ3 > 0.
The detailed definitions of the parameters are given in the mP i Ṗi (t) = uP i (t). (15)
following sequence. Note that λi (·) denotes the ith largest From (13), the control input ωni can be written as
eigenvalue of (·). 
1) Solve XA + AX − 2BB  < 0 to get a solution X > 0. ωni = (uωi + uP i )dt. (16)
Then, R = X −1 .
1
2) Select c1 > a(L) that satisfies S = RA + A R − Remark 1: Note that (14)–(15) and (26)–(27) are the subcases

2c1 a(L)RBB R < 0 while leading to the best of (7) where A = 0, B = 1, and n = 1.
asymptotic consensus. The general algebraic connectivity Remark 2: Considering Remark 1, the system already satis-
a(L) can be found in Definition 2. Select c2 > 0 fies XA + AX − 2BB  < 0. An arbitrary solution of X = 1
that satisfies S̃ = RA + A R − c2 RBB  R < 0, and that is X > 0 is chosen, and in turn, a positive-definite matrix
0 < p ≤ 0.5 that performs a desired finite-time consensus. of R = X −1 = 1 can be determined. Then, K = −B  R = −1
3) Select θ1 , θ 2 , , θ3 > 0 that satisfies ( θc11 + can be decided accordingly.
c2  θ3
θ2 )λmax (RBB R) + 2 < −λmax (Ξ ⊗ S) with Ξ 1) Active Power Sharing: Utilizing (8) and considering Re-
defined in Lemma 2. mark 2, the auxiliary active power control input in (15) is defined
4) Select as
α > c1 θ1 λmax (L Ξ2 L ⊗ RBB  R) uP i (t) = −cP 1 zP i (t) − cP 2 sig(zP i (t) + eP i (t))pP . (17)
2
+ IN ⊗RA−L⊗R(A+c 2θ4
1 BK)

The estimate error eP i (t) and the local regulating error zP i (t)
β > c̄2p (N n)1−p ((L − IN ) ⊗ R2p + 2p L ⊗
for the active power sharing are defined as
R2p ) − ρ λmax (S̃)(L − IN ) ⊗ R2p
γ > 22p c̄(N n)1−p R2p eP i (t) = mP i Pi (t) − mP i P̂i (t) (18)
c̄ = c2 θ2 λmax (M  Ξ2 M ⊗ RBB  R) 
 
+ λmax (LL ⊗R(A+c2θ 1 BK)(A+c1 BK) R)
+ θ24 zP i (t) = aij (mP i P̂i (t) − mP j P̂j (t)) (19)
3
 j∈Ni
+ A R + RA > 0
where θ4 > 0, > 0, 0 < ρ < 1, and M = IN − 1N ξ. where P̂i (t) = Pi (tik ), t ∈ [tik , tik+1 ). The event-triggering time
Note that Ξ and ξ are defined in Lemma 2. tik for agent i is generated when the event-triggering condition
Lemma 2 ([37]): Consider a strongly connected graph fP i (t) ≥ 0 is satisfied. The event-triggering function fP i (t) is
G with Laplacian matrix L. The general algebraic con- defined by

nectivity of G is defined by a(L) = minx ξ,x=0 xx Ξx L̂x
fP i (t, eP i , zP i , ϕP i ) = αP eP i 2 + βP eP i 2pP
where L̂ = (ΞL + L Ξ)/2, Ξ = diag(ξ1 , ξ2 , . . . , ξN ), ξ =
N
[ξ1 , ξ2 , . . . , ξN ] , ξ  L = 0, and i=1 ξi = 1. Note that the
+ γP zP i 2pP − χP εP 1i |ϕP i |2ψP
general algebraic connectivity of an undirected graph is a(L) = (20)
λ2 (L).
Lemma 3 ([34]): If the graph is strongly connected, and the ϕ̇P i (t) = −εP 2i sig(ϕP i (t))ψP . (21)
system (A, B) is controllable, there exist matrices U ∈ C N ×N , The finite-time convergence of the active power sharing using
(N −1)×N
W ∈C  , and Y ∈ C N ×(N −1)
  such that U LU =
−1
(17) and (20) is achieved according to Lemma 1.
0 
0N −1 ξ   
JL = , U −1 = , and U = 1N Y , 2) Frequency Restoration: Similar to the active power shar-
0N −1 Δ W ing, the auxiliary frequency control input (14) is formulated as
where JL is the Jordan canonical form of L and Δ ∈
uωi (t) = −cω1 zωi (t) − cω2 sig(zωi (t) + eωi (t))pω . (22)
C (N −1)×(N −1) is a diagonal matrix.
Lemma 4 ([38]): Consider a continuous nonlinear system The estimate error eωi (t) and the local tracking error zωi (t) for
ẋ = f (x) with f (0) = 0. Suppose there exists a continuously frequency control are respectively defined as
differentiable function V (x) and real numbers a > 0, 0 < b < 1
that satisfy the following conditions: 1) V (x) is positive-definite; eωi (t) = ωi (t) − ω̂i (t) (23)
2) V̇ (x) ≤ −a(V (x))b . Then, the settling time can be upper 
bounded as follows: T (x0 ) ≤ a(1−b) 1
V (x0 )1−b . zωi (t) = aij (ω̂i (t) − ω̂j (t)) + gi (ω̂i (t) − ωref ) (24)
j∈Ni

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 CHOI et al.: DISTRIBUTED FINITE-TIME EVENT-TRIGGERED FREQUENCY AND VOLTAGE CONTROL OF AC MICROGRIDS 1983

where ω̂i (t) = ωi (tik ), t ∈ [tik , tik+1 ). Note from (24) that the 2) Voltage Restoration: Equivalent to the frequency control,
frequency control is a leader-follower problem. In the leader- the auxiliary voltage control input can be written as
follower problem, the pinning gain gi is greater than zero only
uvi (t) = −cv1 zvi (t) − cv2 sig(zvi (t) + evi (t))pv . (34)
if the corresponding ith DG has a direct edge to the leader node,
i.e., DG i receives the reference value ωref . gi equals to zero The estimate error evi (t) and the local tracking error zvi (t) for
otherwise. According to Remark 3, the event-triggering time the voltage control are defined as
for agent i is also determined when fP i (t) ≥ 0. According to
evi (t) = vo,magi (t) − v̂o,magi (t) (35)
Lemma 1, using (20) and (22), the frequency synchronization
can be reached in finite time for any initial conditions as long as 
zvi (t) = aij (v̂o,magi (t) − v̂o,magj (t))
the condition in Lemma 1 is satisfied.
j∈Ni
Remark 3: The frequency and voltage control has faster + gi (v̂o,magi (t) − vref ). (36)
dynamics than the power sharing controls, that is, the stabi-
lization time holds TcP ≈ 10Tcω and TcQ ≈ 10Tcv ; therefore the If the voltage of a critical bus of microgrid is desired to be
frequency/voltage reaches the steady state faster than the ac- synchronized to the nominal voltage, the reference voltage in
tive/reactive power. Generally, the controllers share one common the auxiliary voltage input can be chosen as
communication network, therefore using two separate triggering
vref = kp (vnom − vcrit ) + ki (vnom − vcrit ) (37)
conditions in a different timescale is likely to incur excessive
communication [26]. where vcrit is the critical bus voltage, ki and kp are PI control
gains. According to Lemma 1 and Remark 3, the reactive power
C. FETC Problem for a Secondary Voltage Control of MG sharing and the voltage restoration can be achieved in finite
amount of time using (32) and (33).
Similar to the frequency control, differentiate the reactive
power droop characteristic (2) to obtain the following:
D. Finite-Time Stability
∗
v̇ni (t) = v̇o,magi (t) + nQi Q̇i (t). (25)
The secondary control problem of an islanded AC MG con-
Define the auxiliary voltage and reactive power control inputs, sists of four MAS regulation/tracking problems, namely fre-
uvi and uQi , as follows: quency restoration, active power sharing, voltage restoration,
∗ and reactive power sharing. Since these problems are similarly
v̇o,magi (t) = uvi (t) (26)
structured, herein, the finite-time stability of active power shar-
nQi Q̇i (t) = uQi (t). (27) ing problem is discussed on behalf of secondary control prob-
lems without loss of generality. Denote mP i Pi (t) = Pmi (t) for
By combining (25)–(27), the secondary voltage control input simplicity. The active power sharing problem can be expressed
vni can be defined as as the following matrix form by combining (15), (17)–(19):

vni = (uvi + uQi )dt. (28) Ṗm (t) = − cP 1 L(Pm (t) − eP (t))
− cP 2 sig(L(Pm (t) − eP (t)) + eP (t))pP (38)
1) Reactive Power Sharing: Utilizing (8) and considering
Remark 2, the auxiliary reactive power control input uQi can where Pm (t) = [Pm1 (t), . . . , PmN (t)] and eP (t) =
be written as [eP 1 (t), . . . , eP N (t)] . Define a disagreement vector
uQi (t) = −cQ1 zQi (t) − cQ2 sig(zQi (t) + eQi (t))pQ . (29) ϑP (t) = (IN − 1N ξ  )Pm (t), then its derivative can be
written as
The estimate error eQi(t) and the local tracking error zQi (t) for
the reactive power sharing can be defined as ϑ̇P (t) = − cP 1 L(ϑ(t) − eP (t))

eQi (t) = nQi Qi (t) − nQi Q̂i (t) (30) − cP 2 M sig(L(ϑ(t) − eP (t)) + IN eP (t))pP (39)
 where M = IN − 1N ξ  . Note that M L = L since ξ  L = 0.
zQi (t) = aij (nQi Q̂i (t) − nQj Q̂j (t)). (31) Then, finite-time convergence of the active power sharing prob-
j∈Ni
lem in terms of the disagreement vector is investigated over
Considering Remark 3, the reactive power and voltage mea- microscopic and macroscopic time periods, i.e., convergence
surements can be updated using the event-triggering condition between two events where there is no event triggered in between,
in terms of reactive powers. Utilizing (8) and (12), the event- and convergence over the entire time period [34].
triggering condition fQi (t) can be formulated as Finite-time convergence between events [34]: First, Lyapunov
stability during the communication intervals is discussed. For
fQi (t, eQi , zQi , ϕQi ) = αQ eQi 2 + βQ eQi 2pQ further simplicity, the time term t and subscript P are omitted.
+ γQ zQi 2pQ − χQ εQ1i |ϕQi |2ψQ Select the following Lyapunov function candidate
(32) V = V1 + V2 + V3 (40)
ϕ̇Qi (t) = −εQ2i sig(ϕQi (t))ψQ . (33) V1 = ϑ Ξϑ (41)

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 1984 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 37, NO. 3, MAY 2022

N
 λmax (c21 LL ) θ4
V2 = |qi |p+1 (42) where c̄ = c2 θ2 λmax (M  Ξ2 M ) + + and
1+p 2θ2 2
2p 2p p 2p
i=1 0 < ρ < 1 [34]. Using Lϑ = z + Le ≤ 2 z +
N
 χ 2p L2p e2p , the upper bounds for the third term elements
V3 = |ϕi |1+ψ (43) in (47) are found as follows [34]:
i=1
1+ψ
c̄(sig(q)p ) sig(q)p ≤ c̄2p N 1−p (Lϑ2p + Le − e2p )

where q = L(ϑ − e) + e = [q1 , . . . , qN ] [34]. The derivative
of V1 and its upper bound is found as ≤ c̄22p N 1−p z2p + c̄2p N 1−p ×

V̇1 = ϑ̇ Ξϑ + ϑΞϑ̇ (2p L2p + L − IN 2p ) e2p , (48)

= − 2c1 ϑ ΞLϑ + 2c1 ϑ ΞLe − 2c2 ϑ ΞM sig(q)p ρ λmax (S̃)(sig(q)p ) sig(q)p

≤ λmax (Ξ ⊗ S) ϑ2 ≤ ρ λmax (S̃) Lϑ2p − ρ λmax (S̃) Le − e2p

c1 ≤ ρ λmax (S̃) Lϑ2p − ρ λmax (S̃) L − IN 2p e2p .
+ ϑ2 + c1 θ1 λmax (L Ξ2 L) e2
θ1 (49)
c2
+ ϑ2 + c2 θ2 λmax (M  Ξ2 M ) Then, the inequality (47) can be further bounded by
θ2
c1 c2 θ3
× (sig(q)p ) sig(q)p (44) V̇ ≤ λmax (Ξ ⊗ S) + + + ϑ2
θ1 θ2 2
where θ1 and θ2 are Young’s inequality parameters used to bound c1 L2
V̇1 . Note that Young’s inequality parameters can be any positive + c1 θ1 λmax (L Ξ2 L) + e2
2θ4
numbers that satisfy the condition in Lemma 1. The derivative
of V2 can be found and bounded as
+ c̄2p N 1−p (2p L2p + L − IN 2p )
N
 
V̇2 = sig(qi )p q̇i −ρ λmax (S̃) L − IN 2p e2p
i=1
N

= (sig(q)p ) (−c1 L)ϑ + (sig(q)p ) (c1 L)e  
+ c̄22p N 1−p z2p − χε2i |ϕi |2ψ
+ (sig(q)p ) (S̃)sig(q)p i=1
2p
λmax (c21 L L) θ3 + ρ λmax (S̃) Lϑ
≤ (sig(q)p ) sig(q)p + ϑ2
2θ3 2 + (1 − ρ) λmax (S̃)(sig(q)p ) sig(q)p . (50)
2
c1 L θ4
+ e2 + (sig(q)p ) sig(q)p Assume λmax (Ξ ⊗ S) +
c1
+
c2
+
θ3
< 0. Replace e2 ,
2θ4 2 θ1 θ2 2
+ λmax (S̃)(sig(q)p ) sig(q)p (45) e2p , and z2p terms by (20). Then (50) can be further
bounded by
where θ3 and θ4 are Young’s inequality parameters. Since ϕi is
V̇ ≤ ρ λmax (S̃) Lϑ2p
a predesignated function, the derivative of V3 can be written as
N N
+ (1 − ρ) λmax (S̃)(sig(q)p ) sig(q)p
 
V̇3 = χ sig(ϕi )ψ ϕ̇i = −χε2i |ϕi |2ψ . (46) N

i=1 i=1 + χ(ε1i − ε2i ) |ϕi |2ψ . (51)
i=1
Combining (44)–(46) yields,
In order to further bound each term in (51), the state transforma-
V̇ ≤ λmax (Ξ ⊗ S) +
c1
+
c2
+
θ3
ϑ 2 tion ζ = U −1 ϑ = [ζ1 , ζ2 ] where ζ1 ∈ C and ζ2 ∈ C N −1 , are
θ1 θ2 2 employed [34]. Note that the matrices U , W , Y , and Δ can
be defined from Lemma 3. Using the state transformation, the
c1 L2
+ c1 θ1 λmax (L Ξ2 L) + e2 following can be found:
2θ4  
 ξ
+ c̄ + ρ λmax (S̃) + (1 − ρ) λmax (S̃) ζ= (IN − 1N ξ  )Pm
W
N
 
 0
× (sig(q)p ) sig(q)p − χε2i |ϕi |2ψ (47) = N
Pm (52)
i=1
W (IN − 1N ξ  )

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 CHOI et al.: DISTRIBUTED FINITE-TIME EVENT-TRIGGERED FREQUENCY AND VOLTAGE CONTROL OF AC MICROGRIDS 1985

From (52), ζ1 is found to be zero. Using the fact that ζ1 = may discourage V converging to zero. The value of V2 can be
0, ϑ (L L)ϑ ≥ λmin (U  U )λmin (Δ Δ)ζ2 ζ2 and ϑ Ξϑ ≤ bounded as
λmax (U  ΞU )ζ2 ζ2 , the first, second, and third term in (51) can
be respectively bounded as follows [34]:
 
V2 (tik ) = Lϑ(tik ) − (L − IN )e(tik )p+1
ρ λmax (S̃) Lϑ2p 1+p
p  p+1  p+1
λmin (U  U )λmin (Δ Δ) ≤ α∗ Lϑ(tik ) + β ∗ e(tik ) (59)
≤ ρ λmax (Q̃) V1p
λmax (U  ΞU )

= α V1p , (53) p+1

p+1
where α∗ = 21+p 2
N 2 and β ∗ = α∗ L − IN p+1 [34]. De-
p  p
(1 − ρ) λmax (S̃)(sig(q) ) sig(q) note V2 (tik ) and V2 (t̄ik ) as values of V2 before and after triggering
2p
p+1 2p
at t = tik , respectively. Then, the upper bounds for each case
p+1
≤ (1 − ρ) λmax (Q̃) V2p+1 will be α∗ Lϑ(tik )p+1 + β ∗ e(tik )p+1 and α∗ Lϑ(t̄ik )p+1 +
β ∗ e(t̄ik )p+1 , respectively. It is obvious that the estimate error
2p
is smaller after triggering than before, therefore 0 ≤ e(t̄ik ) ≤
= β V2p+1 , (54)
e(tik ). Also, Lϑ(tik )p+1 = Lϑ(t̄ik )p+1 according to [34].
N
 Therefore, it can be said that the upper bound of V2 is smaller
χ(ε1i − ε2i ) |ϕi |2ψ after triggering than before. For the case where V2 has increased
i=1 after triggering, the maximum magnitude of the increment is
2ψ
2ψ
α∗ Lϑ(t̄ik )p+1 + β ∗ e(t̄ik )p+1 and it can be bounded as
(1 + ψ) 1+ψ
≤ (ε1 − ε2 ) ψ−1 V3ψ+1
χ ψ+1

 p+1  p+1
α∗ Lϑ(t̄ik ) + β ∗ e(t̄ik )
2ψ
ψ+1
= γ V3 . (55)
p+1  p+1  p+1
Note that α , β , γ < 0. Combining (53)–(55), the inequality ≤ α∗ 2 2 z(t̄ik ) + γ ∗ e(t̄ik ) (60)
(51) can be rewritten as
2p 2ψ

V̇ ≤ α V1p + β V2p+1 + γ V3ψ+1 ≤ 0. (56) p+1

where γ ∗ = α∗ 2 2 Lp+1 + β ∗ [34]. From (60), it is observed
2p
From the inequality 0 < ψ ≤ p ≤ 0.5, p < p+1 2ψ
and ψ+1 2p
≤ p+1 that the positive increment of V2 is bounded by z(t̄ik )p+1 and
can be known easily. First, assume the case where 0 < V ≤ e(t̄ik )p+1 which are predestined to reach zero by the threshold
1. Thus, 0 < V1 ≤ 1, 0 < V2 ≤ 1, and 0 < V3 ≤ 1 also hold. (20) and (21). Then, V can converge to zero in finite time based
Then, (56) can be rebounded as on the convergence time discussed for interevent intervals [34].
Combining the analyses for two cases, it can be concluded
2p 2p 2p

V̇ ≤ α V1p+1 + β V2p+1 + γ V3p+1 that V can converge to zero in finite time. In other words,
the active power sharing problem can reach consensus in finite
2p 2p 2p

< max(α , β , γ )(V1p+1 + β V2p+1 + γ V2p+1 ) time utilizing the control protocol (29) and the event-triggering
condition (20)–(21).
2p
< max(α , β , γ )(V1 + V2 + V3 ) p+1
2p
= max(α , β , γ )V p+1 (57)
E. Zeno Behavior Discussion
Applying Lemma 4, V will reach zero in finite-time
Due to the mechanism of event-triggered control, that is,
2p
1− p+1 the communication is triggered when an error term exceeds
V (0)
tf ≤ 2p . (58) the triggering threshold, it is possible that the communication
−max(α , β , γ )(1 − p+1 )
is unlimitedly triggered during a short period of time even if
Now consider a case where V > 1. Based on (56), there exists the error term is small but the threshold is even smaller. The
a positive scalar r such that V̇ ≤ −r. Then, V converges to 1 in aforementioned phenomenon is known as Zeno behavior, and is
finite time [34]. Once it reaches 1, it becomes the case where expected to be analyzed when solving an event-triggered control
0 < V ≤ 1. problem. To prove that Zeno behavior is excluded from the
Finite-time convergence over entire time period [34]: Now, system, the minimum interevent interval needs to be bounded
stability over the entire time period that includes the impact of by a nonzero positive value.
event triggering will be discussed. Note that V1 and V3 are not The Zeno behavior of the proposed FETC strategy is investi-
affected by the communication triggered [34]. V2 is going to gated for two conditions, i.e., when Tc < Tϕ and when Tc ≥ Tϕ ,
have a jump in its value since ei is reset to zero at its trigger where Tc denotes the time when the finite-time consensus is
time t = tik . If the jump in V2 results in the increase of V2 , it reached, and Tϕ is the time when ϕi (t) is converged to zero [34].

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 1986 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 37, NO. 3, MAY 2022

Zeno behavior when Tc < Tϕ : For the active power sharing

control in (15), the estimate error can be written as
 t
eP i (τ ) = uP i (τ )dτ
tik
 t
= (−c1 zP i (τ )
tik

− c2 sig(zP i (τ ) + eP i (τ ))p dτ. (61)

Fig. 3. 4 DG MG test system in Case A & B: (a) circuit diagram; (b)
Then the magnitude of estimate error can be bounded by communication graph.
 t  t
eP i  ≤ c1 zP i  dτ + c2 sig(zP i + eP i )p ) dτ
tik tik
TABLE I
= c1 zP i  (t − tik ) MG SPECIFICATIONS IN CASE A & B

+ c2 sig(zP i + eP i )p ) (t − tik )
= (c1 zP i  + c2 Φ)(t − tik ) (62)
p
where Φ is the upper bound of sig(zP i (t) + eP i (t)) ; Φ can be
bounded because zP i and eP i converge to zero at consensus [34].
Define the upper bound of zP i  as , then c1 zP i  + c2 Φ ≤
c1  + c2 Φ, and let ς = c1  + c2 Φ. Consider (20) and (62),
and let Te be the solution of
ας 2 (Te )2 + βς 2p (Te )2p + γ2p = χε1i |ϕi |2ψ (63)
2 2p 2p 2ψ
where ας , βς , and γ are less than χ 1i |ϕi | [34]. Then,
when an event is triggered, the interevent time interval Te =
tik+1 − tik must be greater than or equal to Te . The solution of
(63) satisfies Te > 0 if ϕi = 0 [34]. Then it can be obtained that
Te ≥ Te > 0 when the threshold does not converge to zero [34].
The above analysis indicates that Zeno behavior is excluded
while ϕi (t) = 0. Therefore for Tc < Tϕ , Zeno behavior is ex-
cluded for the entire time since the consensus is achieved before
ϕi (t) goes to zero.
Zeno behavior when Tc ≥ Tϕ : For this case, Zeno behavior this end, a strategy is proposed to effectively reset the triggering
is excluded for t < Tϕ and t > Tc . Overall, Zeno behavior threshold upon the occurrence of disturbances. In the proposed
is excluded for t ∈ [0, min{Tc , Tϕ }] ∪ [Tc , ∞] for the FETC strategy, the triggering threshold only resets if the error term in
strategy introduced [34]. In fact, the exclusion of Zeno behavior (11) has been less than a certain value for a specific amount of
can be achieved by setting a proper initial value and parameters time before it crosses that value again.
for the threshold such that Tc < Tϕ .

F. Strategy for MG’s Frequency Restoration Subsequent to IV. CASE STUDIES

Disturbances In this section, the effectiveness of the proposed FETC con-
The secondary control is usually applied right after the MG troller is evaluated using two MG test systems. The first MG
islanding. In the proposed FETC scheme, once the islanding includes 4 DGs and is utilized in Cases A and B. The second
is detected, the event-triggering threshold χε1i |ϕi (t)|2ψ in (11) MG that consists of 20 DGs is used in Case C.
adopts it’s initial value and the communication triggering occurs
once the error term in (11) becomes greater than the threshold. In
A. Case A: FETC Verification for Islanded MG With 4 DGs
practical implementation, the triggering is inactivated when the
error term remains smaller than a certain value for some amount Case A validates the effectiveness of FETC control strate-
of time, so that the secondary control can be turned off after gies on a 60 Hz/480 V MG test system with 4 DGs in MAT-
the consensus is achieved. However, upon the occurrence of a LAB/Simulink environment. The single-line diagram of 4 DG
disturbance like load change, the triggering threshold is required MG is shown in Fig. 3(a). The specifications of DGs, lines, and
to reset and readopt it’s initial value to accommodate a limited loads are provided in Table I. The critical bus of MG is DG 4
number of triggering events before the MG distributed frequency bus where Load 2, which is considered as a critical load, is
control restores MG’s frequency to the nominal frequency. To connected.

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 CHOI et al.: DISTRIBUTED FINITE-TIME EVENT-TRIGGERED FREQUENCY AND VOLTAGE CONTROL OF AC MICROGRIDS 1987

TABLE II
CONTROLLER PARAMETERS IN CASE A

Fig. 5. The FETC voltage control and reactive power sharing during MG
islanding event in Case A.1: (a) critical bus voltage; (b) reactive power ratios
nQi Qi .

power ratios of DGs synchronize to an identical value satisfying

(5) and the frequencies are settled around 59.89 Hz by the pri-
mary controller as shown in Fig. 4(a) and (b), respectively. Even
though the primary control is able to settle the DG frequencies,
it cannot avoid the deviation from the nominal frequency of
60 Hz. At t = 2 s, the secondary control is activated and the
Fig. 4. The FETC frequency control and active power sharing during MG frequencies are successfully restored to 60 Hz around t = 4.5 s,
islanding event in Case A.1: (a) DG frequencies; (b) active power ratios mP i Pi . still achieving an accurate active power sharing. For the voltage
control, vref is calculated from (37) with kp = 2 and ki = 4.
Similar to the frequency control, while a proper reactive power
The communication graph for Case A is illustrated in Fig. 3(b) sharing is observed as shown in Fig. 5(b), the critical bus voltage
where each dotted line represents a bidirectional link. The cor- is unable to settle to the nominal value without the secondary
responding adjacency matrix to that graph is control during 0 < t < 2 s as shown in Fig. 5(a). The secondary
⎡ ⎤ control restores the critical bus voltage to 480 V when activated,
0 1 1 0 while obtaining a synchronization of reactive power ratios.
⎢1 0 0 1⎥
⎢ ⎥ Therefore, it can be said that the proposed secondary controller is
A=⎢ ⎥.
⎣1 0 0 1⎦ capable of providing the proper active/reactive power sharings,
0 1 1 0 while simultaneously fulfilling the secondary frequency/voltage
restorations.
The reference frequency and voltage values ωref and vref are sent In order to emphasize the finite-time feature of the proposed
to DG 1 from the leader node, with the pinning gain of g1 = 1. controller, it is compared with the existing asymptotic event-
Every DG is reachable from every other DG, therefore the graph triggered controller presented in [26]. The impact of parameter
is strongly connected. This graph is undirected and its algebraic p in (8) on the finite-time consensus characteristic, is also
connectivity a(L) is 2 that is the second largest eigenvalue of L. discussed. Fig. 6 shows the frequency response of DG 4 to the
The control parameters for Case A are tabulated in Table II. ϕ0 existing asymptotic event-triggered controller in [26] and the
is the initial value of ϕ(t). proposed FETC controller. DG 4 is arbitrarily selected among
1) Case A.1. Islanding From Main Grid: The performance of four DGs. When the secondary control is activated at t = 2 s, the
the proposed secondary controller at the event of MG islanding is proposed FETC controller restores the frequency to 60 Hz with
investigated in this test case. The response of 4 DG MG model more drastic convergence rate than the asymptotic one. It is also
is shown in Figs. 4 and 5. The MG is disconnected from the observed that the frequency is restored faster when p = 0.3 than
main grid at t = 0 s. In order to highlight the effectiveness of p = 0.5. As shown, the finite-time strategy can provide faster
the proposed secondary controller, the secondary control is not restoration rate along with the event-triggered controller, with
applied during 0 < t < 2 s. After the disconnection, the active smaller p providing the faster convergence rate.

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 1988 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 37, NO. 3, MAY 2022

Fig. 6. Frequency response of DG 4 to the existing event-triggered controller

in [26] and FETC controller in Case A.1.

Fig. 8. The impact of ψ on the event-triggered control characteristic in

Case A.1: (a) event-triggering threshold and communication event instances;
(b) frequency of DG 4.

Fig. 7. The impact of ϕ0 on the event-triggered control characteristic in

Case A.1: (a) event-triggering threshold and communication event instances;
(b) frequency of DG 4.

The impacts followed by different values of ϕ0 and ψ on

the event-triggered control characteristic are discussed in this
paragraph. The frequency response and event occurrence of
DG 4 with different ϕ0 are given in Fig. 7. ϕ0 determines the
initial value of a threshold. It can be intuitively thought that the
smaller the threshold is, the more communication is triggered. Fig. 9. The FETC frequency control and active power sharing under load
Fig. 7(a) shows the FETC threshold and event-triggered time for changes in Case A.2: (a) DG frequencies; (b) active power ratios mP i Pi .
ϕ0 = 5, ϕ0 = 10, and ϕ0 = 15. As expected, communication
is triggered the least and the most when ϕ0 = 5 and ϕ0 = 15,
respectively. It is shown in Fig. 7(b) that lesser communications 2) Case A.2. FETC Under Load Changes: An additional
followed by a larger ϕ0 lead to a higher oscillation in frequency amount of P = 4 kW is applied to and removed from Load 1 at
response. The parameter ψ has impact in the curvature of the t = 11 s and t = 15 s, respectively. Fig. 9 shows the frequencies
threshold as shown in Fig. 8(a). Note that ψ affects the initial and active power ratios of DGs. Before the load addition, the
value of the threshold, but in order to observe sole impact of system is stably maintained at 60 Hz while properly distributing
ψ, χ has been adjusted to make the initial threshold values the the active powers among DGs. When load is added at t = 11 s,
same for all cases. As seen, the communication is triggered the the system experiences deviations in frequencies and active
least when ψ = 0.1, and it can be said that smaller ψ decreases power ratios. The proposed controller successfully restores the
chances of triggering as it expands the area where the error term frequencies to 60 Hz and stabilizes the active power sharing
is smaller than the threshold. In Fig. 8(b), it is presented that approximately within 3 s as can be seen in Fig. 9(a) and (b),
sparse communication accompanied by a smaller ψ leads to a respectively. Likewise, when the additional load is removed at
higher oscillation until it reaches consensus. t = 15 s, the system confronts the deviation from the nominal

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 CHOI et al.: DISTRIBUTED FINITE-TIME EVENT-TRIGGERED FREQUENCY AND VOLTAGE CONTROL OF AC MICROGRIDS 1989

Fig. 10. The FETC voltage control and reactive power sharing under load
Fig. 11. The FETC frequency control and active power sharing during DG 3
changes in Case A.2: (a) critical bus voltage; (b) reactive power ratios nQi Qi .
connection and disconnection in Case A.3: (a) DG frequencies; (b) active power
ratios mP i Pi .

states but shortly restored to its desired states by the secondary

controller. Fig. 10 shows the critical bus voltage and reactive
power ratios of DGs. As shown in Fig. 10(a), the proposed
controller can effectively restore the critical bus voltage to 480 V
from the deviation caused by the load changes. After t = 18 s,
DGs’ Reactive powers are shared proportionally based on DGs’
reactive power ratings as shown in Fig. 10(b).
3) Case A.3. Plug-and-Play: DG 3 is disconnected from and
reconnected to the MG at t = 11 s and t = 15 s, respectively.
Its disconnection and re-connection is projected on the network
graph as well. The frequencies and the active power ratios of DGs
are given in Fig. 11. Once DG 3 is disconnected at t = 11 s,
its frequency and active power ratio no longer synchronize
with the rest of DGs. On the other hand, DG 1, DG 2, and
DG 4 are maintained at the consensus state by the proposed
secondary controller even after the disconnection as the graph
is still tied to the leader node through DG 1, and is strongly
connected. After DG 3 is reconnected at t = 15 s, its frequency
and active power ratio are synchronized back to the rest of
Fig. 12. The FETC voltage control and reactive power sharing during DG 3
DGs and the system reaches consensus at 60 Hz by the sec- connection and disconnection in Case A.3: (a) critical bus voltage; (b) reactive
ondary controller. The critical bus voltage control and reactive power ratios nQi Qi .
power sharing are also fulfilled by the secondary controller as
shown in Fig. 12.
4) Case A.4. Resilience Against Communication Link Fail-
ures: A continuous and repeated communication graph changes
are assumed as depicted in Fig. 13. Graphs in Fig. 13 rep-
resent a situation where one or more of the communication
links have failed. Each graph is maintained for 0.05 s and
repeated throughout the simulation. The event of DG islanding
in Case A.1 is simulated under the network graph changes,
and the results are given in Figs. 14 and 15. As shown in Fig. 13. Graph changes every 0.05 s.
the figures, the proposed secondary controllers are capable of

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 1990 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 37, NO. 3, MAY 2022

Fig. 14. The FETC frequency control and active power sharing under com- Fig. 16. The FETC frequency control and active power sharing under com-
munication link failures in Case A.4: (a) DG frequencies; (b) active power ratios munication delays plotted for DG 4 in Case A.5: (a) frequency of DG 4; (b)
mP i P i . active power ratio mP i Pi of DG 4.

Fig. 15. The FETC voltage control and reactive power sharing under commu- Fig. 17. The FETC voltage control and reactive power sharing under commu-
nication link failures in Case A.4: (a) critical bus voltage; (b) reactive power nication delays plotted for DG 4 in Case A.5: (a) critical bus voltage; (b) reactive
ratios nQi Qi . power ratio nQi Qi of DG 4.

fulfilling the frequency/voltage restorations as well as obtaining the proposed controller is still capable of fulfilling the secondary
the active/reactive power sharing. frequency/voltage and active/reactive power sharing goals.
5) Case A.5. Operation Under Communication Delays: In
this test case, the performance of the proposed controller under
the presence of communication delays is evaluated. The system B. Case B: HIL Verification for Islanded MG With 4 DGs
responses at the event of islanding under the delay of 10 ms The 4 DG MG test system model in Fig. 3 that is verified in
and 100 ms are given in Figs. 16 and 17. Such delays are Case A, is tested in the HIL platform in Case B. The control
selected as the communication delays are known to be in range parameters selected for this test case is tabulated in Table III.
of milliseconds to tens of milliseconds [39], [40]. As shown in The HIL platform is implemented as illustrated in Fig. 18.
the figures, the increase in delays result in larger oscillations in Opal-RT is used for a real-time simulator, and MG test system
system response, leading to a slower convergence rate. However, model in Fig. 3 is simulated in Opal-RT with the time step size

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 CHOI et al.: DISTRIBUTED FINITE-TIME EVENT-TRIGGERED FREQUENCY AND VOLTAGE CONTROL OF AC MICROGRIDS 1991

TABLE III
CONTROLLER PARAMETERS IN CASE B

Fig. 19. The FETC frequency control and active power sharing during MG
islanding event in Case B.1: (a) DG frequencies; (b) active power ratios mP i Pi .

Fig. 18. HIL testbed with Opal-RT and Raspberry Pis.

of 100 μs. The proposed secondary controller of each DG is

implemented in an individual Raspberry Pi, accommodated in a
script written in Python. As shown in Fig. 18, there are two
communications required in the implemented HIL platform:
communication between OPAL-RT (MG model) and Raspberry
Pis shown in orange dashed lines, and between Raspberry Pis
in blue dotted lines. Note that the latter represents the network
graph among DGs. Both communications required are realized
by Ethernet; OPAL-RT and the Raspberry Pis are connected to
Ethernet switches under the same network. The data between
each DG and its corresponding controller is transferred via
Modbus TCP/IP. Each DG in the model and each controller
serves as Modbus slave and master, respectively. The communi-
cation graph across controllers is realized on TCP/IP socket. The Fig. 20. The FETC voltage control and reactive power sharing during MG
HIL simulation is processed in the following sequence; with an islanding event in Case B.1: (a) critical bus voltage; (b) reactive power ratios
average measurement sampling and controller updating cycle of nQi Qi .
6.77 ms.
∗
1) The measurements from DG i, ωi , Pi , vo,magi , and Qi , are
sent to the corresponding controller via Modbus TCP/IP. Fig. 19(a), the frequency is settled at 59.9 Hz, deviated from the
2) If the event-triggering condition is satisfied, the controller nominal frequency of 60 Hz. The secondary frequency control
∗
i transmits ωi , Pi , vo,magi , and Qi to its out-neighbor is activated at t = 5 s, and the frequency is restored to 60 Hz
controllers via TCP/IP socket, i.e., controllers j that are around t = 25 s. Also, Fig. 19(b) shows that the active power
∗
aji > 0; and controller i receives ωj , Pj , vo,magj , and Qj ratios of all DGs are synchronized to each other before and after
from its in-neighbors, i.e., controllers j that are aij > 0. the secondary control is applied, implying that the active power
3) The controller i then calculates ωni and vni using (16) and sharing is achieved through the primary control. For the voltage
(2), and sends ωni and vni to DG i in the model through control, kp = 1 and ki = 0.3 are used to calculate vref . As shown
Modbus TCP/IP. in Fig. 20(a), the critical bus voltage is settled around 471 V
1) Case B.1. Islanding From Main Grid: The performance deviated from 480 V, when the secondary control is not activated.
of the proposed controller when MG is islanded, is evaluated in When the secondary voltage controller is applied at t = 5 s,
this test case. The results are given in Figs. 19 and 20. Initially, the critical bus voltage is restored to 480 V around t = 25 s.
MG is in an islanded mode and the system is stabilized without Fig. 20(b) shows that the proposed controller can also provide
the secondary control applied during 0 ≤ t ≤ 5 s. As shown in the reactive power sharing.

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 1992 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 37, NO. 3, MAY 2022

Fig. 21. The FETC frequency control and active power sharing under load
changes in Case B.2: (a) DG frequencies; (b) active power ratios mP i Pi . Fig. 22. The FETC voltage control and reactive power sharing under load
changes in Case B.2: (a) critical bus voltage; (b) reactive power ratios nQi Qi .

The average event intervals for the frequency/active power

and voltage/reactive power sharing control during the transient
period 5 ≤ t ≤ 25, are 12.3 ms and 10.6 ms, respectively,
whereas it would be 7 ms without the FETC strategy. Therefore it
is verified from the HIL test, that the FETC strategy can provide
frequency/voltage restoration and active/reactive power sharings
in finite-time convergence rate, while reducing the burden in
communication network.
2) Case B.2. FETC Under Load Changes: An additional
load P = 4 kW is applied to and removed from DG 1 bus at
t = 80 s and t = 120 s, respectively. The frequencies and the
active power ratios of DGs are given in Fig. 21. When the
load is added or removed, the frequency deviates from 60 Hz
and the synchronization of active power ratios is interrupted.
However, the frequency is restored to 60 Hz around t = 95 Hz
and the active power ratios are synchronized again. The critical
bus voltage and the reactive power ratios are shown in Fig. 22.
Likewise, the critical bus voltage is restored to 480 V from
deviation caused by the load changes, while providing a proper
reactive power sharing.

C. Case c: Model Verification for Islanded MG With 20 DGs

In order to test the scalability of the proposed controller, it is
applied to a 20 DG MG test system illustrated in Fig. 23(a).
The specifications of the MG model are listed in Table IV.
The communication network for this test case, is designed as
illustrated in Fig. 23(b). DG 1 has a direct edge from the leader
node with the pinning gain of g1 = 1. The algebraic connectivity
of the given graph is a(L) = 5.6180. The FETC controller and
the event-triggering function are configured by the parameters
provided in Table V. The output frequencies and the active power
Fig. 23. 20 DG MG test system in Case C: (a) circuit diagram; (b) communi-
ratios of the DGs under the given control configuration, are cation graph.
shown in Fig. 24. In Fig. 24(a) after the disconnection from
the main grid at t = 0, the frequencies start to diverge from

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 CHOI et al.: DISTRIBUTED FINITE-TIME EVENT-TRIGGERED FREQUENCY AND VOLTAGE CONTROL OF AC MICROGRIDS 1993

TABLE IV
MG SPECIFICATIONS IN CASE C

Fig. 25. Communication triggered time of DGs in Case C.

V. CONCLUSION
This paper proposes a FETC control strategy for the secondary
frequency and voltage regulation of the distributed MGs. The
proposed strategy controls MG frequency and voltage while
sharing the active and reactive power among the DGs based on
their ratings. The proposed FETC control provides the system
consensus in finite-time as well as reducing the communication
energy compared to the continuous communication controller.
TABLE V The model simulation results using 4 DG MG test system,
CONTROLLER PARAMETERS IN CASE C
suggest the efficacy of the proposed controller, supporting with
various test cases: response at an islanding event, load changes,
plug-and-play, communication link failures, and communication
delays. The HIL simulation results further strengthen the effec-
tiveness of the proposed strategy with two case studies: islanding
event and load changes. The scalability of the proposed con-
troller is also verified through the simulation in a 20 DG test MG
system. The future research directions will include investigating
the cyber security of FETC secondary control of microgrids and
creating attack resilient FETC control techniques.

REFERENCES
[1] J. M. Guerrero, J. C. Vasquez, J. Matas, L. G. De Vicuña, and M. Castilla,
“Hierarchical control of droop-controlled AC and DC microgrids-A gen-
eral approach toward standardization,” IEEE Trans. Ind. Electron., vol. 58,
no. 1, pp. 158–172, Jan. 2011.
[2] A. Bidram and A. Davoudi, “Hierarchical structure of microgrids control
system,” IEEE Trans. Smart Grid, vol. 3, no. 4, pp. 1963–1976, Dec. 2012.
[3] A. Bidram, A. Davoudi, F. L. Lewis, and J. M. Guerrero, “Distributed
cooperative secondary control of microgrids using feedback linearization,”
IEEE Trans. Power Syst., vol. 28, no. 3, pp. 3462–3470, Aug. 2013.
[4] A. Bidram, A. Davoudi, and F. L. Lewis, “A multiobjective distributed con-
trol framework for islanded AC microgrids,” IEEE Trans. Ind. Informat.,
vol. 10, no. 3, pp. 1785–1798, Aug. 2014.
[5] Q. Shafiee, J. M. Guerrero, and J. C. Vasquez, “Distributed secondary
control for islanded microgrids-A novel approach,” IEEE Trans. Power
Electron., vol. 29, no. 2, pp. 1018–1031, Feb. 2014.
Fig. 24. The FETC frequency control and active power sharing in Case C: (a) [6] J. W. Simpson-Porco, F. Dörfler, and F. Bullo, “Synchronization and power
DG frequencies; (b) active power ratios mP i Pi . sharing for droop-controlled inverters in islanded microgrids,” Automatica,
vol. 49, no. 9, pp. 2603–2611, Sep. 2013.
[7] R. Heydari, M. S. Golsorkhi, M. Savaghebi, T. Dragicevic, and F.
Blaabjerg, “Communication-free secondary frequency and voltage control
of VSC-based microgrids: A high-bandwidth approach,” in Proc. 22nd Eur.
the nominal frequency of 60 Hz. The secondary control acts Conf. Power Electron. Appl., 2020, pp. 1–9.
at t = 1 s and the frequency is restored to 60 Hz at t = 5 s, [8] S. Li, H. Du, and X. Lin, “Finite-time consensus algorithm for multi-agent
while providing the successful active power sharing as shown in systems with double-integrator dynamics,” Automatica, vol. 47, no. 8,
pp. 1706–1712, Aug. 2011.
Fig. 24(b). The communication triggering timestamps for each [9] X. Lu, R. Lu, S. Chen, and J. Lu, “Finite-time distributed tracking control
DG are shown in Fig. 25. The average communication interval of for multi-agent systems with a virtual leader,” IEEE Trans. Circuits Syst.
the proposed FETC-based control during 1 < t < 5 s, is 0.19 ms. I., Reg. Papers, vol. 60, no. 2, pp. 352–362, Feb. 2013.

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.
 1994 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 37, NO. 3, MAY 2022

[10] Z.-H. Guan, F.-L. Sun, Y.-W. Wang, and T. Li, “Finite-time consensus Proc. IEEE 3rd Adv. Inf. Technol., Electron. Automat. Control Conf., 2018,
for leader-following second-order multi-agent networks,” IEEE Trans. pp. 1974–1978.
Circuits Syst. I., Reg. Papers, vol. 59, no. 11, pp. 2646–2654, Nov. 2012. [33] L. Zhang, Z. Zhang, N. Lawrance, J. Nieto, and R. Siegwart, “Decen-
[11] J. Bickel, “An overview of transients in power systems,” tralised finite-time consensus for second-order multi-agent system un-
2019, [Online]. Available: https://download.schneider-electric. der event-triggered strategy,” IET Control Theory Appl., vol. 14, no. 4,
com/files?p_enDocType=WhitePaper&p_File_Name=998- pp. 664–673, Mar. 2020.
20579579_GMA.pdf&p_Doc_Ref=998-20579579_GMA. [34] C. Du, X. Liu, W. Ren, P. Lu, and H. Liu, “Finite-time consensus for
[12] A. Bidram, A. Davoudi, and F. Lewis, “Finite-time frequency synchroniza- linear multi-agent systems via event-triggered strategy without continuous
tion in microgrids,” in Proc. IEEE Energy Convers. Congr. Expo., 2014, communication,” IEEE Trans. Control Netw. Syst., vol. 7, no. 1, pp. 19–29,
pp. 2648–2654. Mar. 2020.
[13] F. Guo, C. Wen, J. Mao, and Y.-D. Song, “Distributed secondary volt- [35] A. Bidram, F. L. Lewis, and A. Davoudi, “Synchronization of nonlinear
age and frequency restoration control of droop-controlled inverter-based heterogeneous cooperative systems using input-output feedback lineariza-
microgrids,” IEEE Trans. Ind. Electron., vol. 62, no. 7, pp. 4355–4364, tion,” Automatica, vol. 50, no. 10, pp. 2578–2585, 2014.
Jul. 2015. [36] A. Bidram, V. Nasirian, A. Davoudi, and F. L. Lewis, Cooperative Synchro-
[14] S. Zuo, A. Davoudi, Y. Song, and F. L. Lewis, “Distributed finite-time nization in Distributed Microgrid Control. Cham, Switzerland: Springer,
voltage and frequency restoration in islanded AC microgrids,” IEEE Trans. 2017.
Ind. Electron., vol. 63, no. 10, pp. 5988–5997, Oct. 2016. [37] W. Yu, G. Chen, M. Cao, and J. Kurths, “Second-order consensus for
[15] N. M. Dehkordi, N. Sadati, and M. Hamzeh, “Distributed robust finite-time multiagent systems with directed topologies and nonlinear dynamics,”
secondary voltage and frequency control of islanded microgrids,” IEEE IEEE Trans. Syst., Man, Cybern. B., Cybern., vol. 40, no. 3, pp. 881–891,
Trans. Power Syst., vol. 32, no. 5, pp. 3648–3659, Sep. 2017. Jun. 2010.
[16] D. V. Dimarogonas, E. Frazzoli, and K. H. Johansson, “Distributed event- [38] S. P. Bhat and D. S. Bernstein, “Finite-time stability of continuous au-
triggered control for multi-agent systems,” IEEE Trans. Autom. Control, tonomous systems,” SIAM J. Control Optim., vol. 38, no. 3, pp. 751–766,
vol. 57, no. 5, pp. 1291–1297, May 2012. 2000.
[17] E. Garcia, Y. Cao, and D. W. Casbeer, “Decentralized event-triggered [39] Q. Shafiee, Č. Stefanović, T. Dragičević, P. Popovski, J. C. Vasquez, and J.
consensus with general linear dynamics,” Automatica, vol. 50, no. 10, M. Guerrero, “Robust networked control scheme for distributed secondary
pp. 2633–2640, Oct. 2014. control of islanded microgrids,” IEEE Tran. Ind. Electron., vol. 61, no. 10,
[18] C. Nowzari and J. Cortés, “Zeno-free, distributed event-triggered commu- pp. 5363–5374, Oct. 2014.
nication and control for multi-agent average consensus,” in Proc. Amer. [40] C. Ahumada, R. Cárdenas, D. Saez, and J. M. Guerrero, “Secondary
Control Conf., 2014, pp. 2148–2153. control strategies for frequency restoration in islanded microgrids with
[19] Y. Fan, G. Hu, and M. Egerstedt, “Distributed reactive power sharing consideration of communication delays,” IEEE Trans. Smart Grid, vol. 7,
control for microgrids with event-triggered communication,” IEEE Trans. no. 3, pp. 1430–1441, Aug. 2015.
Control Syst. Technol., vol. 25, no. 1, pp. 118–128, Jan. 2017.
[20] M. Chen, X. Xiao, and J. M. Guerrero, “Secondary restoration control of Jeewon Choi (Student Member, IEEE) received the
islanded microgrids with a decentralized event-triggered strategy,” IEEE B.S. degree in mechanical and automotive engineer-
Trans. Ind. Informat., vol. 14, no. 9, pp. 3870–3880, Sep. 2018. ing from Keimyung University, Daegu, South Korea,
[21] L. Ding, Q.-L. Han, and X.-M. Zhang, “Distributed secondary control in 2015, and the M.S. degree in 2018 in mechanical
for active power sharing and frequency regulation in islanded microgrids engineering from the University of New Mexico, Al-
using an event-triggered communication mechanism,” IEEE Trans. Ind. buquerque, NM, USA, where she is currently working
Informat., vol. 15, no. 7, pp. 3910–3922, Jul. 2019. toward the Ph.D. degree in mechanical engineering.
[22] S. Weng, D. Yue, C. Dou, J. Shi, and C. Huang, “Distributed event-triggered Her research interests include smart grid, microgrid,
cooperative control for frequency and voltage stability and power sharing and cyber-physical system modeling and simulation.
in isolated inverter-based microgrid,” IEEE Trans. Cybern., vol. 49, no. 4,
pp. 1427–1439, Apr. 2019.
[23] T. Qian, Y. Liu, W. Zhang, W. Tang, and M. Shahidehpour, “Event- Seyed Iman Habibi (Graduate Student Member,
triggered updating method in centralized and distributed secondary con- IEEE) received the B.Sc. and M.Sc. degrees in elec-
trols for islanded microgrid restoration,” IEEE Trans. Smart Grid, vol. 11, trical engineering from the K. N. Toosi University
no. 2, pp. 1387–1395, Mar. 2020. of Technology, Tehran, Iran, in 2015 and 2018, re-
[24] B. Abdolmaleki, Q. Shafiee, M. M. Arefi, and T. Dragičević, “An instanta- spectively. He is currently working toward the Ph.D.
neous event-triggered Hz-watt control for microgrids,” IEEE Trans. Power degree with the University of New Mexico, Albu-
Syst., vol. 34, no. 5, pp. 3616–3625, Sep. 2019. querque, NM, USA. His main research interests in-
[25] B. Abdolmaleki, Q. Shafiee, A. R. Seifi, M. M. Arefi, and F. Blaabjerg, clude secondary control of microgrids, supervisory
“A zeno-free event-triggered secondary control for AC microgrids,” IEEE control, switching control, and adaptive systems.
Trans. Smart Grid, vol. 11, no. 3, pp. 1905–1916, May 2020.
[26] Y. Wang, T. L. Nguyen, Y. Xu, Z. Li, Q.-T. Tran, and R. Caire, “Cyber-
physical design and implementation of distributed event-triggered sec- Ali Bidram (Senior Member, IEEE) received the
ondary control in islanded microgrids,” IEEE Trans. Ind. Appl., vol. 55, B.Sc. and M.Sc. degrees from the Isfahan Univer-
no. 6, pp. 5631–5642, Nov. 2019. sity of Technology, Isfahan, Iran, in 2008 and 2010,
[27] Y. Zhu, X. Guan, X. Luo, and S. Li, “Finite-time consensus of multi-agent respectively, and the Ph.D. degree from the Univer-
system via nonlinear event-triggered control strategy,” IET Control Theory sity of Texas at Arlington, Arlington, TX, USA, in
Appl., vol. 9, no. 17, pp. 2548–2552, Nov. 2015. 2014. He is currently an Assistant Professor with the
[28] H. Zhang, D. Yue, X. Yin, S. Hu, and C. xia Dou, “Finite-time dis- Electrical and Computer Engineering Department,
tributed event-triggered consensus control for multi-agent systems,” Inf. University of New Mexico, Albuquerque, NM, USA.
Sci., vol. 339, pp. 132–142, Apr. 2016. Before joining University of New Mexico, he was
[29] Q. Lu, Q.-L. Han, B. Zhang, D. Liu, and S. Liu, “Cooperative control with Quanta Technology, LLC, and was involved in a
of mobile sensor networks for environmental monitoring: An event- wide range of projects in electric power industry. He is
triggered finite-time control scheme,” IEEE Trans. Cybern., vol. 47, no. 12, an Associate Editor for the IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS.
pp. 4134–4147, Dec. 2017. His area of expertise lies within control and coordination of energy assets
[30] Y. Dong and J.-g. Xian, “Finite-time event-triggered consensus for non- in power electronics-intensive energy distribution grids. Such research efforts
linear multi-agent networks under directed network topology,” IET Control are culminated in a book, several journal papers in top publication venues and
Theory Appl., vol. 11, no. 15, pp. 2458–2464, Oct. 2017. articles in peer-reviewed conference proceedings, and technical reports. He
[31] Z. Cao, C. Li, X. Wang, and T. Huang, “Finite-time consensus of linear was the recipient of the IEEE Albuquerque Section Outstanding Engineering
multi-agent system via distributed event-triggered strategy,” J. Franklin Educator Award, New Mexico EPSCoR Mentorship Award, University of Texas
Inst., vol. 355, no. 3, pp. 1338–1350, Feb. 2018. at Arlington N. M. Stelmakh Outstanding Student Research Award, Quanta
[32] J. Li, X. Wu, Q. Li, and Y. Li, “Finite-time distributed event-triggered con- Technology Shooting Star Award, IEEE Kansas Power and Energy Conference
sensus control for leader-following general linear multi-agent systems,” in Best Paper Award, and cover article of December 2014 in IEEE Control Systems.

Authorized licensed use limited to: M S RAMAIAH INSTITUTE OF TECHNOLOGY. Downloaded on October 20,2023 at 05:35:55 UTC from IEEE Xplore. Restrictions apply.